A DARPA Sarcasm Detector Is Good

We may earn a commission from links on this page.
Image for article titled A DARPA Sarcasm Detector Is Good
Photo: Denis Charlet (Getty Images)

Members of the media are sarcastically complaining about a sarcasm detector developed by researchers partnering with the Defense Advanced Research Projects Agency (DARPA). Using sarcastic language, they indicate that it is a stupid killjoy only useful to GOP politicians. Before eye-rolling on Twitter, hear me out. An optional sarcasm detection tool could do a profound public service.

DARPA announced earlier this month that researchers at the University of Central Florida have developed AI that picks up on “cue-words” as part of DARPA’s Computational Simulation of Online Social Behavior (SocialSim) program. They claim that the model, tested on headlines from the Onion and the Huffington Post, pulled off a “nearly perfect sarcasm detection score” on a dataset of tweets. Sarcasm pertains to national security, they explain, because it could help understand bad actors’ undermining of fact and promote crucial information during disasters.

“Sarcasm has been a major hurdle to increasing the accuracy of sentiment analysis, especially on social media, since sarcasm relies heavily on vocal tones, facial expressions, and gestures that cannot be represented in text,” Brian Kettler, a program manager in DARPA’s Information Innovation Office, is quoted in the presser. “Recognizing sarcasm in textual online communication is no easy task as none of these cues are readily available.” It doesn’t suggest imposing government labels, nor do I have any idea how that would work, but it’s not a bad idea for a Twitter feature.


As decades of studies have shown, some people on the autism spectrum struggle to detect figurative language. Researchers have found that, while people with ASDs often possess above-average intelligence, children with ASDs can have difficulty in grasping metaphors, ambiguity, and idioms compared to neurotypical people. One of the cited theories is that this is generally a problem when receiving information from multiple sources. Rolling takes are more or less the purpose of Twitter.

This is the subject of research by Annuska Zolyomi, an incoming teaching professor at the University of Washington Bothell, who co-authored a study on autism-oriented Twitter conversations. “Users discuss the challenge in sensing jokes and sarcasm, especially in the written modality,” the study reads. “One user shared that ‘entire threads here on Twitter can make zero sense to me because it’s in the form of sarcasm.’” (Hence the need for the marker “/S.”) Some users, the study notes, would like to see Twitter add something like an opt-in “joke” button.


“A lot of my research is about reframing social communication practices, both online and in-person,” Zolyomi told Gizmodo. “It’s not the autistic person who should change or adapt to fit neurotypical social norms. It’s a social group as a whole that should work together to have mutual understanding and establish common ground.”

Emphasis on “mutual.” Zack Budryk, a reporter at The Hill who has written about being on the autism spectrum and runs a Twitter feed containing masterful ableism satire, told Gizmodo that a sarcasm tool could help neurotypical followers get it. “I think it could help neurotypical people understand the kind of extremely deadpan sarcasm that I have found to be fairly common among autistic people but that NT people may miss,” he wrote over Twitter.


“I think it could be extremely helpful to some people on the spectrum,” he said, “but I also think the thing is that a lot of us sort of grow up extremely sarcastic ourselves as a defense mechanism against those difficulties. (Greta Thunberg’s tweets are a great example). So if anything, this could make it easier for autistic people’s use of sarcasm to be recognized as such.”

It’s also unfair to frame this solely as an issue for people on the spectrum. Morons consistently struggle with parody.


Separately, other groups struggle to grasp that infinitesimal probabilities might as well be zero, construing a 0.00001% chance of a negative outcome with “a chance”–which correlates with anxiety-related disorder tendencies to exaggerate the probability of negative outcomes, demonstrated in people with flying phobia and claustrophobia.

This is why reports of huge asteroids that frequently populate The Sun, The Mirror, and primarily British read tabloid The Daily Express might as well be called the Daily Asteroid Heart Attack Broadcast with a salvo of headlines such as “Asteroid Shock: ‘Hand out Bibles’” and “Asteroid danger: 100% certainty of impact warns space expert.” They’re not sarcastic statements, but exaggerated headlines make for easy parody.


Taking everything on its face would make the internet a living nightmare. Hyperbolic headlines are so damaging to some people that a debunker named Robert Walker has devoted his life to interpreting them for an over 2,000-member Facebook group Doomsday Debunked. Even a hyperbolic claim that a piece of legislation will “end the internet” stokes real fear for some Doomsday Debunked members.

A sarcasm detector would not ruin the internet or its relentless flow of jokes. We will always have dril.